Operating Matrix/PV Scoring Study

Quote from laziz314159:

Ahh, but that bit isn't trend, it's just acc/dis.

example:

bar 0 is positive close, 1.5k vol, bar 1 is even close, 3k vol, bar 2 is negative close, 2k vol. If you just carried the sign forward, acc/dis wouldn't tick down until after bar2, whereas the present construction ticks down on bar2.

I think carrying the sign foward would synchronise the jma(8) and acc/dis too much. It's all an inkblot test anyway. ;)

Thanks for your comments,
Laz
More...

Yes, I see your point, again. And, I can understand plotting acc/dis as a binary. Let us think one step further: would it be possible that what you might be needing is to link these two up? If they are both going up at the same time -- that's green. If they are going different directions -- that's blue. If they are both going down at the same time -- that's red.
 
Quote from bdixon619:

Yes, I see your point, again. And, I can understand plotting acc/dis as a binary. Let us think one step further: would it be possible that what you might be needing is to link these two up? If they are both going up at the same time -- that's green. If they are going different directions -- that's blue. If they are both going down at the same time -- that's red.
More...

Sorry, didn't mean to repeat myself. PV links all three together (I guess the name's a tad misleading, PVAD might be better):

solution sets for the combos you noted:

(5,7), (1,3,4,6), (0,2) in pvScore
i.e.
(101,111), (001,011,100,110), (000,010)

 
Thank you. I find it interesting that 2/3 are rejection patterns and 1/3 continuation patterns. Looking closely at your histograms in the Attachment provides this clue. The ones I call 'green' and 'red' are basically the same patterns. The last thing I've seen is pattern 4 bears out what the chart shows: volume spikes at the end of downtrends.
 
I took the text file Laziz provided and created an input file for my rule generator. I have attached the rules in a text file and added a picture of the decision surface.

There are two input variables, R1 and R2, and one output variable, Output. The file Laziz provided was a single column of numbers from 0 to 7 corresponding to the market states as described in his paper. What I did was to create two more columns of the same information in the same order. Since, there are eight market states (0-7) and two Input variables, 64 Rules are created when each of the R1 Input variables is matched with all 8 of the R2 input variables. These inputs are then compared with the third column of numbers (Output).

Since Laziz provided over 4,000 records spanning a few days we can build a short term statistical picture of the trading environment. That is what is represented in the Decision Surface. R1 and R2 are shown on the X and Y axis, while Output is plotted on the Z (vertical) axis. The higher the peak the more frequently a certain R1-R2 pair results in a certain output.

This picture and description differs from Laziz' report in just one way: it adds one level of complexity. Looking at Laziz' histograms we are looking at the 'Output' of an 8 Rule rulebase...the size of the different bins in the histogram reflects how each market state moves to some other market state. Some states tend to move more frequently in certain directions than other states.

This was an interesting time.
 

Attachments

this is one of the most interesting threads on ET......although I am not a futures trader and not that much into the construction of such a model to trade from........this is for sure some deep thought.

I hope you guys can put this into BLACKBOX form and make ur end product work.

Best of luck and thanks for the intellectual Insight.
 
While this was a fun exercise and was a good learning experience, I want to mention here that this was a short term model. It should be pretty clear that a model like this is only a sample of the total possibilities available as solutions to the question of what comes next in the market.

If a guy or gal were to start with two months' worth of tick data and then update the model every night, then that would probably be what it would take to use this system in intraday trading.

I've thought instead that possibly it would be better to use daily data and get a year or so of that and then use the daily predicted output as a kind of bias as to how you may want to trade in the market for any given day.

I am really greatful Laziz took the time to decipher what Jack was saying. I haven't any idea if Jack had made this leap himself or simply had observed enough to provide the energy that Laziz was able to use to make the leap of intuition. Certainly, Laziz deserves credit for a masterful job of data organization.
 
When we're passing the praise around here, let's not forget about Jack. He's the man that got this started.

Then again, there was probably somebody behind him - probably a woman! LOL

(note: i posted this before seeing bdixon's previous post)

chaos
 
You must log in or register to reply here.
Back
Top